20220802, 15:56  #45 
Jul 2004
Milan, Ita
2·107 Posts 
ftree file from stage 2 overloads trash in Colab
Started testing 30.9 b1 on a couple of Colab instances on 13M PRPCF exponents (10 curves, B1=50000, B2=15000000) and noticed soon a large <exp>.ftree file (~1.2GB) during stage 2 of each curve.
When stage 2 is completed, the <exp>.ftree file gets deleted and moved to the Trash folder  exhausting soon the 15GB space limit. Once proper debugging is done, would it be possible to truncate that file to zero or minimallength, prior to deletion? 
20220802, 16:12  #46  
P90 years forever!
Aug 2002
Yeehaw, FL
1111101101000_{2} Posts 
Quote:
In the meantime, set EcmSaveFtreeToDisk=0 in prime.txt 

20220809, 11:08  #47 
"Tony Gott"
Aug 2002
Yell, Shetland, UK
101100101_{2} Posts 
mprime?
In amongst everything else would you be willing to upgrade mprime to 30.9 so I could run ECM, please ...

20220809, 11:53  #48 
"TF79LL86GIMPS96gpu17"
Mar 2017
US midwest
3×2,293 Posts 
George already has posted both Windows and Linux builds
at https://mersenneforum.org/showpost.p...4&postcount=10 in this thread, and pointed to by both https://www.mersenneforum.org/showpo...35&postcount=1 (Announcements) and https://www.mersenneforum.org/showpo...77&postcount=1 (prime95/mprime intro) in the reference info. Last fiddled with by kriesel on 20220809 at 11:58 
20220809, 12:12  #49  
"Tony Gott"
Aug 2002
Yell, Shetland, UK
3×7×17 Posts 
Quote:


20220812, 01:40  #50 
"James Heinrich"
May 2004
exNorthern Ontario
3,851 Posts 
Nothing specific to v30.9, problem has likely existed for a long time, but a user pointed out to me that Prime95 reports TFF results incompletely:
{"status":"F", "exponent":512792989, "worktype":"TF", "factors":["10621798844143574859209"], (etc) What's missing is bitlo, bithi, rangecomplete keys, it should look like this: {"status":"F", "exponent":512792989, "worktype":"TF", "factors":["10621798844143574859209"], "bitlo":73, "bithi":74, "rangecomplete":false, (etc) TFNF results should still include the same 3 keys with "rangecomplete":true. 
20220816, 03:11  #51 
"Cong Shengzhuo"
Sep 2021
Nanjing, China
37 Posts 
Does v30.9 mean faster ECM?
Personally, I'm quite determined to see if we can reach <200 unfactored exponents per 10K range for small exponents, and the first range with >=200 unfactored is 2.22M. But the P1 bounds of 2.22M is high enough, and there is no need to reP1. (See this factoring effort page.) So I am hoping if someone can help doing ECM in the range with the benefits from v30.9. Last fiddled with by congsz on 20220816 at 03:12 
20220816, 04:37  #52  
Dec 2021
3^{3} Posts 
Quote:


20220816, 04:47  #53 
P90 years forever!
Aug 2002
Yeehaw, FL
17550_{8} Posts 

20220816, 05:10  #54  
"Cong Shengzhuo"
Sep 2021
Nanjing, China
37 Posts 
Quote:
I don't quite understand the mechanism of ECM. If I want to reach t35 on the exponent, should I pick B1=250000, complete 640 curves, and then pick B1=1000000 and complete 1566 curves? Or just choose B1=1000000 from the start of my work? Another question: Is the ECM B2 in v30.9 determined automatically (like P1 B2 in v30.8) or manually? If manually, how would I choose B2? 

20220816, 05:26  #55 
P90 years forever!
Aug 2002
Yeehaw, FL
17550_{8} Posts 
Like P1, prime95 ECM will choose B2 automatically. To simplify matters I'd just start with B1 = 1000000. Depending the B2 chosen by prime95 will determine the number of curves that need to be run. It will be much less than 1566.
Last fiddled with by Prime95 on 20220816 at 05:26 
Thread Tools  
Similar Threads  
Thread  Thread Starter  Forum  Replies  Last Post 
That's a Lot of Users!!!  jinydu  Lounge  9  20061110 00:14 
Beta version 24.6  Athlon users wanted  Prime95  Software  139  20050330 12:13 
For Old Users  Citrix  Prime Sierpinski Project  15  20040822 16:43 
Opportunity! Retaining new users postM40  GP2  Lounge  55  20031121 21:08 
AMD USERS  ET_  Lounge  3  20031011 16:52 